Leveraging Optimized FFT on Intel Xeon Platforms

Help Desk

Video

This is an interactive videoconference on leveraging Fast Fourier Transform (FFT) for Intel Xeon Platforms. FFT is a performance-critical component of many HPC applications, from signal processing to cosmological simulations. An N-dimensional FFT is decomposed into 1D FFT of complex-to-complex and real-to-complex and data transposes. Highly optimized FFT libraries, such as FFTW and MKL FFT, provide the building blocks for specialized FFT implementations. This talk discusses the use patterns arising from real applications, the factors that are critical for the performance, and how to use the basic FFT library to maximize the performance on Intel Xeon and Xeon Phi clusters.

Presenter
Jeongnim Kim, Intel